Файл: [D._Kim_Rossmo]_Geographic_Profiling(BookFi.org).pdf

© 2000 by CRC Press LLC

variables to predict MTC: R3 rapist type. They found interesting results with
the predictive domains of expressive aggression (e.g., injuries requiring med-
ical treatment, cuts, bruises, abrasions, biting, weapon use), and adult anti-
social behaviour (e.g., alcohol use, drug use, presence of weapon). Those
domains with larger numbers of variables had better prediction power.

Evidence of the ability to predict rapist type from crime-scene variables

provides an empirical framework for profiling. Soley et al. suggest that “geo-
graphic profiling technology … could be joined with crime-scene personality
profiling (CIA) to provide law enforcement with immediate, empirically-
based information not only on rapist type, but also on where the rapist may
be operating from” (p. 18). To some extent the Royal Newfoundland Con-
stabulary (RNC) Criminal Suspect Prioritization System (CSPS) does this
through the use of offence themes to determine the likely criminal background
of a rapist or armed robber, and then applying this information in conjunction
with geographic proximity to identify potential offenders from existing police
records (House, 1997). Valid and reliable offender classification schemes ame-
nable to quantification and capture in a computer database are first necessary
for this approach to become standard police investigative practice.

Promising research in Britain by Davies, Wittebrood, and Jackson (1997,

1998) demonstrates that certain crime scene behaviours exhibited by stranger
rapists are useful for predicting elements of the offender’s criminal record.
Variables such as concealment of identity, familiarity with the criminal justice
system, control of the victim, method of approach, criminal behaviour, and
alcohol involvement, were used in a logistic regression to determine criminal
record characteristics. Their model best predicted prior convictions for bur-
glary (69% accurate), violent crime (59% accurate), and one-off crimes (i.e.,
no other sexual offences) (71% accurate). These models are potentially valuable
tools in the generation and prioritization of suspects in sexual assault cases.

An FBI study of 41 incarcerated serial rapists responsible for a minimum

of 10 victims begins to provide a more detailed description of this type of
sex offender (Hazelwood & Warren, 1995b). The rapists in the sample
attacked a total of 837 victims, an average of 20.4 victims per offender. The
majority of the victims were strangers (84%), and half were attacked in their
own homes, usually while alone (79%). In contrast to popular perception,
only 12% of these rapes occurred outside (6% in streets or alleys, and 6% in
parking lots or on highways). Warr (1988) estimates that 50 to 60% of all
rapes are residential. These findings challenge the assumption of rape as a
street crime, and the perception of safety within the home.

In many ways these sexual offenders appeared normal. The majority had

stable employment, lived with someone (78%), and had been married at least
once (71%). Most resided in single-family dwellings, but a significant minor-
ity lived in apartments. They tested at above average intelligence. But a

© 2000 by CRC Press LLC

significant number had problems as juveniles. The study noted the presence
of enuresis (32%), cruelty to animals (19%), and fire setting (24%) — the
Macdonald triad — in the sample. Also common were stealing and shoplift-
ing (71%), youthful alcohol abuse (63%), and assaults against adults (55%).
The majority began their careers as voyeurs (68%). For example, the Ski
Mask Rapist, Jon Berry Simonis, began peeping when he was 15 years of age,
before progressing to exposing, obscene telephone calls, and finally rape
(Michaud & Hazelwood, 1998).

There was evidence of abuse in the backgrounds of most of these offend-

ers; 76% claimed to have been sexually victimized as youngsters, 73%
psychological abuse, and 38% physical abuse. Most of the time it was a parent
or caretaker responsible for the abuse. In those instances of sexual molesta-
tion originating from a family member, the person responsible was just as
likely to be female as male, but if the abuse came from a stranger, the
victimizer was almost always male.

The mean age at first rape was 21.8 years, at middle rape, 25.8 years, and

at last rape, 29 years. Almost all of this group had prior arrests and 58% had
been institutionalized (46% in correctional centres and 12% in mental health
facilities). Previous crimes commonly included residential burglaries close to
the offender’s home. Although they had been convicted of a mean of 7.6
previous sexual assaults, they reported actually being responsible for a mean
of 27.8 such crimes. In other words, they were convicted of only one sex
crime out of every 3.7 they committed.

Grubin and Gunn (1990) found that 86% of the English and Welsh serial

and single rapists (n = 142) they studied had a criminal history, typically
involving some type of theft. Half had four or more previous convictions,
and 29% received their first sentence before 20 years of age. There was a
greater percentage of serial rapists with previous sex crimes, usually indecent
exposure or indecent assault, than single rapists (46 versus 25%). The
research noted a pattern of increasing criminality amongst rapists.

The FBI research observed premeditated rape to be more prevalent than

opportunistic or impulsive rape, and it is not uncommon for victims to be
selected through offenders window peeping or following women home. Prior
to the attack, several rapists entered the victim’s home during her absence to
gain familiarity with the premises. Despite this, Hazelwood and Warren
(1995b) caution that most serial rapists do not hunt or stalk particular
individual victims. Rather, their hunt appears to be, like that of most offend-
ers, haphazard and only roughly preplanned; about one-third of the offenders
had consumed alcohol prior to their crimes, and disguises were only worn
in 7 to 12% of the cases.

Hazelwood and Warren (1995b) conclude that target selection in serial

rape is typically not symbolic. Instead, victims are chosen because of general

© 2000 by CRC Press LLC

proximity, availability, and premise access. Victim selection criteria reported
by the offenders in this study include availability (98%), gender (95%),
location (66%), age (66%), race (63%), physical characteristics (39%), cloth-
ing (15%), and no specific traits (25%).

15

 The average victim was in her

twenties, but 19% were children. The research found that white serial rapists,
unlike black rapists, do not cross race lines. White European males were
observed to be the most sexually deviant offenders.

Vehicles were commonly used during the rapes; of these, 62% were the

offender’s, 7% were borrowed, and 8% were the victim’s. No stolen vehicles
were involved in the crimes associated with this study. Some rapists drove
their victims to cemeteries to commit the assault because they believed that
police rarely patrol graveyards (Michaud & Hazelwood, 1998).

The surprise attack was the most commonly used approach, followed by

the con and then the blitz. Rapists stay twice as long on average with victims
who resist. Following an attack, 12 to 15% revisited the crime scene, 8 to
13% communicated with the victim, and 28% followed the investigation in
the media. About half (44 to 51%) of the offenders felt guilty afterwards.

Other research conducted by the FBI suggests that five crime scene behav-

iours can be used to predict “increasers,” rapists who increase their use of
violence in future rapes (Hazelwood, Reboussin, & Warren, 1989; Warren et
al., 1991): (1) no negotiation with the victim; (2) lack of victim reassurance;
(3) use of bindings; (4) transportation of the victim from the encounter site;
and (5) macho offender image. This scale correctly predicted violence esca-
lation in 89% of the research cases. Increasers comprised 25% of the studied
rapist population. On average, they had twice as many victims (40 versus
22), and their attacks were three times as frequent as non-increasers (a mean
of 19 days versus 55 days between crimes).

These findings were not replicated by either Grubin and Gunn (1990)

or by Warren et al. (1995). The former analysis of serial rapists from England
and Wales (n = 11) found that increasers tended to be younger and suffer
from premature ejaculation. The latter study of U.S. serial rapists (n = 108)
observed that white rapists were more likely to increase their level of violence.
The inconsistencies between the studies may be attributable to sample dif-
ferences in level of criminal activity.

In another FBI study of serial rape, the concept of the convex hull polygon

(CHP) was used to characterize and examine spatial patterns of serial rapists.
The CHP is the area enclosed by the convex polygon that connects the outer
locations of a crime series, containing by definition all the offender’s crime
sites. The residence of the rapist was located within the CHP in 24%, and
outside the CHP in 76% of the cases. Warren et al. (1995) created four

15 

Percentages do not add to 100 as multiple responses were allowed.

© 2000 by CRC Press LLC

geographical models based on two dichotomized parameters — offender
residence (within or outside the CHP), and mean distance to crime sites
(near or far). The frequencies for these groupings in their data were as follows:
(1) model 1 (inside near), 13%; (2) model 2 (inside far), 11%; (3) model 3
(outside near), 37%; and (4) model 4 (outside far), 39%.

The typical crime in this research involved a stranger victim (92%),

attacked by surprise (78%), inside their home (60.2%), during the early
morning hours (32% between 18:00 and midnight, 52% between midnight
and 06:00) (Warren et al., 1995). Anger excitation rapists showed a somewhat
different pattern. They were more likely to exhibit specific victim selection
criteria (55% versus 16%), and almost all (91%) committed their crimes in
the victim’s home.

Certain crime scene aspects help identify components of offender spatial

behaviour. For example, the FBI serial rape study looked at differences
between marauders and commuters. The former commit their crimes around
home, while the latter commute to a different region to offend (these concepts
are discussed later in more detail). Nighttime offending by serial rapists was
found to be a modal characteristic for 90% of the marauders, compared to
70% of the commuters (Warren et al., 1995). This result could stem from
marauders wishing to protect their identities, or simply from commuter
travel logistics. Commuters also demonstrated more ritualized behaviour
than marauders. Minimum crime trip distances (i.e., smaller “nonoffending
areas”) were shorter for marauders than for commuters (0.74 versus 2.51
miles), nonwhites than whites (1.23 versus 2.70 miles), younger than older
offenders (0.59 versus 2.58 miles), and nighttime than daytime rapists (1.44
versus 3.12 miles). Not surprisingly, minimum and mean crime trip distances
were correlated.

Other findings of geographic interest include:

•

61% of the rapes occurred indoors;

•

The most common initial contact site was the victim’s home (60.2%),
followed by a public street (20.4%);

•

51% of the rapists were commuters; and

•

A vehicle was used in only 15% of the assaults.

Hazelwood and Warren (forthcoming) found ritual to be a determinant

of sex offender spatial behaviour.  They classify serial sexual criminals into
the impulsive offender, who commits unsophisticated, reactive, and
unplanned crimes, and the less common ritualistic offender, who commits
planned, rehearsed, and generally more sophisticated crimes.  Impulsive sex
offenders, characterized by diverse criminal histories and generic sexual inter-
ests, are usually less specific in victim selection and not as careful regarding

© 2000 by CRC Press LLC

identification than ritualistic offenders, who have pervasive fantasies and
diverse paraphilic interests.  Consequently, the former tend to travel shorter
distances to offend (2.30 versus 3.64 miles), and commit their crimes over
smaller areas (4.57 versus 20.39 mi

2

).

Experienced rapists prefer to attack women in their homes than on the

street, because once they are inside they are relatively safe from observation or
interference (see Warr, 1988). Hazelwood recounts the case of an educated power
reassurance rapist who preselected up to six potential victims. If the first rape
attempt failed, he assaulted a backup victim, her home serving as his “safe house”
while police responded to the first attack. It is not surprising the most common
nonsexual offence in a rapist’s background is breaking and entering. Some of
their crimes are actually “bonus rapes,” the result of an offender opportunisti-
cally encountering a woman during a burglary (Warr, 1988).

16

 Also, many

attempted sexual assaults involving forcible entry are demoted to burglary; if
police cannot prove the former, they will charge the offender with what the
available evidence supports (Michaud & Hazelwood, 1998). Conversely, fetish
and voyeuristic burglaries may be precursors to sexual assaults and homicides.
Schlesinger and Revitch (1999) observed that 42.3% of the sexual murderers
(n = 52) they studied possessed a history of burglary.

In his study of 30 serial rapists in British Columbia, responsible for 183

incidents, Alston (1994) determined rape locations were spatially patterned.
Kolmogorov-Smirnov tests for goodness of fit revealed significant clustering
and the presence of distance decay in the data. Serial rapists used a narrow
search space that closely followed their activity paths. Using curvilinear dis-
tance

17

 measures, Alston found that 94% of the offences were within 2.5

kilometres of the offender’s activity space, and most were within 0.30 kilo-
metres. Activity space was defined as the offender’s activity nodes (past and
present homes, current and previous work sites, and residences of partners,
friends, and family members) and connecting routine pathways. When activ-
ity nodes alone were considered, 72% of the offences were within 5 kilome-
tres, and most were within 2 kilometres. The mean curvimetre distance from
rape site to nearest offender activity node was 0.91 kilometres (median =
0.72 km, standard deviation = 0.89). The mean curvimetre distance from
rape site to nearest offender routine pathway was 0.53 kilometres (median =
0.41 km, standard deviation = 0.93).

For those crimes where the geographical milieu was known, prostitution

strolls and hitchhiking regions appeared to be favoured target regions for

16 

Felson and Clarke (1998) note that one type of crime can generate opportunities for

another type of crime. Some rapes are byproducts of burglary, robbery, or other property
offences.

17 

Curvilinear (curvimetre) distance is the “wheel distance,” or the length measured fol-

lowing a street network. Such distances are usually longer (they can be equal to, but never
shorter) than straight-line crow-flight measures.